---
title: Apple PPO limitations and how practitioners work around them
slug: ppo-limitations-and-workarounds
summary: The practical constraints of Apple's Product Page Optimization (new releases kill tests, review delays, one test at a time, aggregated international data, icon tests need a binary) and the workarounds Popcore and Klarna used in 2022–2023.
domain: mobile-apps
tags: [aso, product-page-optimization, ab-testing, app-store-connect, icons]
related: [app-store-screenshot-production-and-ppo-setup, aso-ab-test-sample-size-and-bold-variants, aso-ab-test-reading-results-and-ties, custom-product-pages-for-ads]
sources: [yt:-YsOukrqcWU, yt:lW0kO3KvyFE, yt:3nas91cJjDE, yt:M2e6pJuA8hA]
updated: 2026-10-24
---

# Apple PPO limitations and how practitioners work around them

## TL;DR
- **Releasing a new app version stops a running PPO test** (2022–23). Plan tests around the release roadmap; a win needs ~7 days running + ~2 days for analytics to catch up.[^1][^2]
- **Screenshot, video and icon tests don't need App Review**; other treatments do. Klarna, with weekly releases, tests **screenshot reorders** right after each release to get a full 7 days.[^3][^4]
- Stuck "Waiting for review" >2 days? **Cancel and resubmit** — usually passes in hours. Avoid submitting on Fridays.[^5][^6]
- **One test at a time**, and multi-locale tests report **aggregated** country metrics — Popcore tests en-US first, then 1–2 top locales, then rolls the winner to small locales untested.[^7][^8]
- **Icon tests need every icon in the binary** (Xcode 13+ asset catalog), and applying a winner needs another binary release.[^9][^10]

## Details
Sources date from PPO's first two years: Marina Roglic (Popcore, ASO Conference 2022), Mateusz Wrzeszcz (Klarna, 2023), Jevgen Tarasenko (Tilting Point, Feb 2022), Simon Thillay (AppTweak, Feb 2022). PPO launched 7 Dec 2021.[^11]

**Release conflicts.** Marina's workaround is organisational: get on the product team's release channel and roadmap, sell the business case (native, free, winner is one click away), and share wins publicly so production becomes willing to wait.[^2][^12] Mateusz could only run real iOS tests during a three-week release freeze.[^13] Storemaven's Jonathan Fishman (2022) saw no company delaying a release for a screenshot test, which is why PPO lagged Google Play experiments in adoption.[^14][^15]

**Review and upload quirks (Marina, 2022).**
- Drafting a test, then starting it after the version ships fails — the tested version shows as expired.[^3]
- Assets stuck "processing": delete and re-upload.[^6]
- In PPO the video poster frame doesn't keep the timestamp you set; she places the intended poster frame at second 29 of the video, which works 9 times out of 10.[^7]

**Localisation strategy.** Because results can't be split per locale and analytics show countries, not localisations, Popcore borrows a head start from Android learnings, then sequences: default language → top "whale" locales → apply elsewhere and monitor KPIs.[^7][^8]

**Icon tests.** Treatments (up to three plus default) must ship in the binary and pass review; the test itself then needs no review. Applying the winner can't be done remotely — stop, ship a new binary with it as default, wait for review again. Popcore lacked dev capacity and used an external tool instead.[^9][^10]

**CPPs aren't a substitute.** Some developers reported success replacing PPO with custom product pages; Popcore did not.[^16]

## Disagreements / caveats
- These are 2022–2023 observations. The 2026 notes in this base ([[app-store-screenshot-production-and-ppo-setup]], Steve P. Young) still describe one draft at a time and submit-for-review, but don't confirm whether a new release still stops a test — check current App Store Connect behaviour.
- Jevgen (2022) called native testing "revolutionary" despite these issues; Simon said PPO had so many implementation issues that it lacked even the appeal of Google Play experiments.[^17][^14]

## See also
- [[aso-ab-test-sample-size-and-bold-variants]]
- [[aso-ab-test-reading-results-and-ties]] — re-analysing PPO data with a Bayesian calculator

[^1]: yt:-YsOukrqcWU @ 04:03 — Marina Roglic
[^2]: yt:-YsOukrqcWU @ 04:59 — Marina Roglic
[^3]: yt:-YsOukrqcWU @ 08:09 — Marina Roglic
[^4]: yt:lW0kO3KvyFE @ 38:06 — Mateusz Wrzeszcz
[^5]: yt:-YsOukrqcWU @ 08:58 — Marina Roglic
[^6]: yt:-YsOukrqcWU @ 09:48 — Marina Roglic
[^7]: yt:-YsOukrqcWU @ 10:36 — Marina Roglic
[^8]: yt:-YsOukrqcWU @ 11:24 — Marina Roglic
[^9]: yt:-YsOukrqcWU @ 14:33 — Marina Roglic
[^10]: yt:-YsOukrqcWU @ 15:29 — Marina Roglic
[^11]: yt:-YsOukrqcWU @ 01:34 — Marina Roglic
[^12]: yt:-YsOukrqcWU @ 05:45 — Marina Roglic
[^13]: yt:lW0kO3KvyFE @ 36:17 — Mateusz Wrzeszcz
[^14]: yt:M2e6pJuA8hA @ 29:21 — Simon Thillay
[^15]: yt:M2e6pJuA8hA @ 33:53 — Jonathan Fishman
[^16]: yt:-YsOukrqcWU @ 16:22 — Marina Roglic
[^17]: yt:3nas91cJjDE @ 13:17 — Jevgen Tarasenko